改进Canny算子的列车轮对踏面边缘检测算法  

Edge Detection Algorithm of Wheelset Tread Based on Improved Canny Operator

作　　者：刘成刚 刘二林[1] 姜香菊[2] LIU Chenggang;LIU Erlin;JIANG Xiangju(School of Mechanical and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070 Gansu,China;School of Automation and Electrial Engineering,Lanzhou Jiaotong University,Lanzhou 730070 Gansu,China)

机构地区：[1]兰州交通大学机电工程学院,兰州730070 [2]兰州交通大学自动化与电气工程学院,兰州730070

出　　处：《铁道机车车辆》2021年第1期22-25,共4页Railway Locomotive & Car

摘　　要：现有对高速列车轮对踏面损伤检测逐步以动态检测系统代替人工检测,针对动态检测系统中对踏面损伤图像处理时,存在大量边缘信息丢失、图像细节不完整等问题,提出一种基于改进Canny算子的边缘检测算法。该算法利用自适应双边滤波代替传统高斯滤波对图像进行平滑处理,能够很好的保持图像边缘和滤除噪声;采用水平、垂直、45°和135°4个方向梯度模板来确定梯度幅值;用Otsu算法(大津法)自适应地根据图像灰度值生成高阈值和低阈值,克服人为设定高低阈值的缺陷,增强了算法的自适应性。通过仿真实验结果表明:改进的算法能够有效的保留更多的踏面图像边缘信息,也具有较强的自适应性,为轮对踏面损伤动态检测的后处理提供了有效的技术支持。At present,dynamic detection system is gradually substituting for manual detection for the damage detection of high-speed train wheelset tread.Aiming at the problems of large amount of edge information loss and incomplete image details in the image processing of tread damage,proposing an improved edge detection algorithm bases on Canny operator.This algorithm uses adaptive bilateral filtering in⁃stead of traditional Gaussian filtering to smooth the image which can maintain the image edge and remove the noise well.The gradient ampli⁃tude was determined by using horizontal,vertical 45°and 135°gradient templates.Otsu algorithm(Dajin method)is used to generate high threshold and low threshold according to the image gray value adaptively,so as to overcome the defect of setting high threshold and low threshold artificially and enhance the self-adaptability of the algorithm.Simulation results show that the improved algorithm can effectively re⁃tain more tread image edge information and has strong self-adaptability,thus providing effective technical support for the post-processing of dynamic damage detection of wheelset tread.

关 键 词：轮对踏面 损伤 CANNY算子 边缘检测 双边滤波 

分 类 号：U260.111[机械工程—车辆工程]